All use cases

Prevent / Care

Injection-site reaction triage

Your patient finds a red, itchy lump where they injected and isn't sure if they did it wrong or if something is wrong — the agent fixes the technique that causes most of it, and gets a clinician on the rest.

Next best thingGet aheadBehavioral triggerRisk triggerChatWhatsAppPushIn-appGLP-1 & PeptidesHealth & Telehealth

What it does

Most people on a self-injected medication will see redness, itching, or a small lump at the site at some point, and almost none of it is dangerous — but the patient can't tell a normal reaction from an infection, so the harmless case quietly erodes their confidence and the rare serious one waits in the same queue. The agent meets the report where it lands, in chat or on a photo, separates the technique problem from the safety problem, and handles each correctly: protocol-safe fixes for the everyday reaction, an immediate clinician hand-off when the picture looks like infection. It reads the situation and routes it; it never diagnoses and never tells the patient to change a dose.

How it works

  1. 1
    Trigger. a patient reports or photographs an injection-site reaction — burning, redness, itching, or a lump — in a conversation or on a check-in.
  2. 2
    Decision. the agent reads the description or image against the patient's medication, recent dose, and prior reports, then branches on two questions — is this fixable technique, or does it carry an infection red flag (spreading bright redness, blistering, pus, lasting several days)? Conservative defaults: anything ambiguous escalates.
  3. 3
    Action. for an ordinary reaction it shares the program's approved technique fixes — let a refrigerated dose reach room temperature, rotate sites, keep the area clean and dry, use a cold compress — and what to watch for; for a red-flag picture it routes straight to a clinician with the report and history attached, and stands automated guidance down on that thread.
  4. 4
    Follow-through. the reaction, the read, and the routing are logged to the record; where guidance was given, the agent checks back to confirm it settled, and any "it's getting worse" reply re-triages and escalates at once.
Example
Patientthere's a hard red lump where I did my shot last night and it itches, is that bad?
AgentThat itchy lump a day after injecting is one of the most common reactions and is usually minor. A few things help: let the pen sit out for 15–30 minutes so it isn't cold going in, move to a fresh spot each week, and a cold compress eases the itch. Please send me a photo or message me if it turns bright red and spreads, blisters, leaks fluid, or is still there in a few days — I'll get a clinician to look right away. I'll check in with you tomorrow.

Configuration

How the agent is wired for this use case.

Triggeran injection-site reaction report or image in a patient conversation (reaction_reported / photo_received) on chat, WhatsApp, in-app, or a scheduled check-in.
Tools & actions
  • Messaging channel · receive the report or photo and deliver approved technique guidance in the flow of conversation.
  • Knowledge base · retrieve the program's injection-technique fixes and the infection red-flag / watch-for criteria.
  • Clinical/EHR system · read the patient's medication, recent dose, and prior site reports to weigh the report in context.
  • Clinician escalation queue · hand off any red-flag or ambiguous reaction with the report and history attached, and stand automated replies down on that thread.
  • Scheduling system · set a check-back and re-triage on any worsening reply.
Autonomyreport intake, technique guidance from approved content, and the check-back run unattended under judge gating. The agent reads apparent severity and routes only — no diagnosis, no dosing or schedule advice, under a clinical-safety policy (detect → escalate). Any infection red flag or ambiguous picture is a mandatory clinician hand-off.
Channelschat · whatsapp · push · in-app
Escalationa red-flag visual or described feature (spreading bright redness, blistering, pus, multi-day persistence), a worsening follow-up reply, or any request for a diagnosis or dose change routes immediately to a clinician with the report and patient context attached.

What you need

The inputs this use case runs on. Your channels stay yours; the agent supplies the judgment.

Signals

injection-site reaction reports and inbound images in patient conversations, optional accompanying text, and follow-up replies.

Data

medication and recent dose, prior site reports and photos, persistent patient memory, consent for image handling, the program's approved technique guidance and red-flag criteria.

Guardrails

the agent assesses severity and routes only — no diagnosis, dosing, or schedule advice ever; conservative defaults with mandatory clinician hand-off on any red-flag or ambiguous reaction; explicit consent and secure, compliant handling of medical images; judge gating on every message; a clear scope boundary stated to the patient; full audit trail of every triage decision.

Metrics it moves

  • contact-ratedown: the everyday "is this lump normal?" reaction resolves with safe technique guidance instead of consuming a clinical slot.
  • time-to-resolutiondown: an infection-pattern reaction reaches a clinician the moment it's reported rather than waiting in a shared photo queue.
  • adherenceup: patients who get the technique right and feel watched over keep injecting instead of skipping doses over a sore site.

Related use cases

Prevent / Care

Photo-based symptom triage and routing

the generic visual-symptom router this specializes to the injection site

Prevent / Care

GLP-1 side-effect education with red-flag escalation

the any-symptom education-and-escalate flow this sits beside

Onboard / Activate

First-injection onboarding and technique coach

where correct injection technique is first taught

See it on your own customer journey

Bring one drop-off, one churn cliff, or one silent segment. We will show you what a proactive agent with memory and judgment does with it.

Book a demo